目的：评价不同操作者使用视网膜自动图像分析(automated retinal image analyzer，ARIA)软件测量糖尿病患者视网膜血管管径和迂曲度的重复性和再现性。方法：采用前瞻性诊断性试验研究设计，由2名熟练操作者应用ARIA软件对49名糖尿病患者以视盘为中心的45°数码眼底彩照进行视网膜血管半自动分析。通过测量距离视盘0.5~1.0个视盘直径内的视网膜动静脉血管参数，各选取6条最粗的动静脉综合计算得到中央动脉管径当量(central retinal artery equivalent，CRAE)、视网膜中央静脉管径当量(central retinal vein equivalent，CRVE)、视网膜动静脉比值(arteriole-to-venule ratio，AVR)、平均视网膜动脉迂曲度(mean retinal arteriolar tortuosity，MRAT)、平均视网膜静脉迂曲度(mean retinal venular tortuosity，MRVT)。在此基础上，评价同一操作者与不同操作者间测量结果的重复性和再现性，并绘制Bland-Altman图进行一致性分析。结果：操作者A两次测量的CRAE、CRVE、AVR差异无统计学意义(t= -0.089、-1.425、1.214，均P>0.05)，重复性良好(ICC≥0.725)且高度正相关(r≥0.721)。操作者A与操作者B对同一受试者2次测量的CRAE、CRVE、AVR差异无统计学意义(t= -1.169、-0.050、0.941，均P>0.05)，再现性良好(ICC≥0.750)且高度正相关( r≥0.697)。同一操作者和不同操作者间的MR AT、MRVT差异无统计学意义

(P≥0.05)，Spearman非参数相关性分析重复性和再现性较好(r≥0.571，r≥0.588，均P<0.01)。结论：利用ARIA软件测量糖尿病患者视网膜血管管径及迂曲度具有良好的重复性和再现性。

Objective: To assess the repeatability and reproducibility in measurements of retinal vessel diameter and tortuosity in diabetic patients using the Automated Retinal Image Analyzer (ARIA). Methods: A prospective diagnostic trial was performed; 2 skilled researchers performed a semi-automatic analysis of retinal vessels in digital color fundus photography (45° field of view centered on the optic disc) from 49 diabetic patients using ARIA software. Six largest arterioles and venules coursing through an area of 0.5–1.0 disc diameter surrounding the optic disc were selected. Then the retinal vessel parameters were calculated into the central retinal artery equivalent (CRAE), central retinal vein equivalent (CRVE), arteriole-to-venule ratio (AVR), mean retinal arteriolar tortuosity (MRAT) and mean retinal venular tortuosity (MRVT). To evaluate the repeatability of those values in 2 measurements by the same researcher and the reproducibility between different researchers. Bland-Altman analysis was employed to assess the agreement of the outcomes. Results: There was no significant difference in the CRAE, CRVE, and AVR values in the two-time measurements of researcher A (t= -0.089, -1.425 and 1.214, all P>0.05). It also showed a good repeatability and a highly positive correlation (ICC≥0.725, r≥0.721). There was no significant difference in those measurements between researcher A and B (t= -1.169, -0.050, 0.941, all P>0.05). It also showed a good reproducibility and a highly positive correlation (ICC≥0.750, r≥0.697). Results showed that there was no significant difference in the MRAT and MRVT values both in intraobserver and interobserver measurements (P≥0.05). Besides, Spearman’s nonparametric correlation analysis showed good repeatability and reproducibility (r≥0.571, r≥0.588, all P<0.01). Conclusion: ARIA software yields good repeatability and reproducibility in the measurements of retinal vessel diameter and tortuosity in diabetic patients.